van der Schaar Lab

Automated machine learning as a partner in predictive modelling

Our new article: Automated machine learning can redefine clinical risk prediction. It can empower both modelling experts and non-experts, democratize access to machine learning methods, and encode higher standards in model development.

In the Lancet Digital Health, we argue that by democratising access to state-of-the-art techniques and encoding good practice to improve the quality of models, automated machine learning frameworks are likely to play a central role in the future of clinical risk prediction.

The traditional approach to developing clinical risk prediction models is largely subject to the expertise of the developer(s). Fast-forward this to machine learning and the technical challenge becomes harder: more algorithms, more hyper-parameters, more code. This has contributed to the poor quality of many ML models proposed in clinical medicine.

In fact, the technical challenge of tuning machine learning algorithms is such that some estimates have it that only 5% of time in machine learning model development is spent on actual modelling questions rather than programming. This should be inverted: the focus should be on solving risk prediction problems well, augmenting experts.

Instead of relying on developers to understand how to effectively tune all state-of-the-art ML approaches, we should use software.

We already have such software. It can efficiently select and train machine learning pipelines using any statistical or machine learning algorithm, performing a task that is presently impractical, if not impossible, even for those with substantial expertise. This means we can always have a high-quality benchmark model for any given dataset against which other approaches can be trialled. And we can see if and where ML provides advantages and where it is unnecessary.

Automation does more though. It can encode good practice upfront, rather than having to rely on post-hoc checklist-based methods to evaluate work at the point of publication.

Our recommendations for key principles and recommendations of automated machine learning frameworks in healthcare are:

Thomas Callender

Mihaela van der Schaar

Mihaela van der Schaar is the John Humphrey Plummer Professor of Machine Learning, Artificial Intelligence and Medicine at the University of Cambridge and a Fellow at The Alan Turing Institute in London.

Mihaela has received numerous awards, including the Oon Prize on Preventative Medicine from the University of Cambridge (2018), a National Science Foundation CAREER Award (2004), 3 IBM Faculty Awards, the IBM Exploratory Stream Analytics Innovation Award, the Philips Make a Difference Award and several best paper awards, including the IEEE Darlington Award.

In 2019, she was identified by National Endowment for Science, Technology and the Arts as the most-cited female AI researcher in the UK. She was also elected as a 2019 “Star in Computer Networking and Communications” by N²Women. Her research expertise span signal and image processing, communication networks, network science, multimedia, game theory, distributed systems, machine learning and AI.

Mihaela’s research focus is on machine learning, AI and operations research for healthcare and medicine.